Re: [PATCH] gpio: Document the 'valid_mask' being internal

[Matti Vaittinen]

Hi dee Ho peeps (Biju, Geert, Linus and all)

On 28/02/2025 12:28, Biju Das wrote:
> Hi Geert,
>
> > -----Original Message-----
> > From: Geert Uytterhoeven <geert@xxxxxxxxxxxxxx>
> > On Fri, 28 Feb 2025 at 09:07, Linus Walleij <linus.walleij@xxxxxxxxxx> wrote:
> > > On Wed, Feb 26, 2025 at 12:42 PM Matti Vaittinen
> > > <mazziesaccount@xxxxxxxxx> wrote:
> > > > On 26/02/2025 12:18, Linus Walleij wrote:


> I agree, when the code is mainlined at that time set_multiple() has some draw backs and hence
> the check is added to take care of GPIO holes.

If I don't read it wrong, rcar GPIO supports some input enabling "en 
masse" during the probe. It seems to me the gpio_rcar_enable_inputs() 
does also need the valid GPIOs information - I suppose some of the GPIOs 
may have been masked in the device-tree, and those shouldn't be enabled.

It feels counter productive to hide the valid_mask - and do some 
device-tree parsing in the driver(s) which may need it.

I suppose we can still hide the valid_mask in struct gpio_device as 
suggested - but then we should probably create a getter for it in the 
gpiolib.

Or does someone see a way around needing the valid_mask in the 
gpio_rcar_enable_inputs() ?

Have a nice weekend!

Yours,
	-- Matti